CS Colloquium

Time and Location:

Speaker:

Link:

Algorithms for understanding 3D protein structure are essential for basic biology, drug discovery, and biotechnology. My research develops deep generative models, particularly diffusion models, for predicting, simulating, and designing protein structure, going beyond predictive modeling paradigms popularized by AlphaFold2. This talk will be organized around three areas where my algorithmic contributions have enabled important modeling applications. First, by extending standard diffusion models to the non-Euclidean manifold of ligand poses, I will describe the development of a state-of-the-art system for protein-ligand docking, a key task in computational drug discovery. Second, by adapting AlphaFold to a bespoke flow-matching framework, I will describe a transferable system for emulating protein dynamics at significantly reduced cost compared to molecular dynamics simulations. Finally, by constructing a multi-objective optimization problem with predictive models, I will describe a workflow for designing proteins with complex design specifications, for which training data is scarce or unavailable. I will conclude by envisioning how a continued synergy of algorithmic innovations and appreciation of biological applications can accelerate the maturation of protein design as a systematic engineering discipline.

Bowen is a Ph.D. candidate in Electrical Engineering and Computer Science at MIT, co-advised by Tommi Jaakkola and Bonnie Berger. His research focuses on generative models for scientific applications in drug discovery, molecular simulation, and protein design. His work has been adopted by multiple drug discovery software providers and recognized with 4 spotlight / oral presentations at NeurIPS / ICLR / ICML. Previously, Bowen completed a B.S. in Computer Science at Stanford with Ron Dror and developed geometric vector perceptrons, one of the most widely-used neural network architectures for protein structure.

In-person attendance only available to those with active NYU ID cards.